The reviewed record of science sign in
Pith

arxiv: 2607.06449 · v1 · pith:WH5N4V4Y · submitted 2026-07-07 · math.NA · cs.NA

Sparse space-time spectral methods can time-step by peel and pass

Timon S. Gutleb This is my paper

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel 2026-07-08 05:12 UTCglm-5.2pith:WH5N4V4Yrecord.jsonopen to challenge →

classification math.NA cs.NA MSC 65M7065M1265M6041A10
keywords space-time spectral methodstime steppingJacobi polynomialsLegendre polynomialsChebyshev polynomialsspectral elementsorthogonal polynomialsendpoint identity
0
0 comments X

The pith

Spectral time-stepping via coefficient summation

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper identifies a property of certain orthogonal polynomial bases (Legendre and Chebyshev first-kind) that makes block-by-block time-stepping possible for space-time spectral methods. In these endpoint-benign bases, every basis polynomial equals one at the right endpoint, so the final-time solution of a solved block is recovered exactly by summing the stored coefficients along the time index. This peel-and-pass step, which the author traces to a Jacobi endpoint identity, lets one solve a space-time spectral problem on a short time block, extract the final slice in coefficient space with no interpolation or quadrature error, and feed it as the initial condition for the next block. The result is a spectral element method in time that holds only one block in memory at a time, needs fewer time coefficients per block than a single global solve would need for the full interval, and preserves spectral accuracy. The paper proves the underlying identities, derives memory, cost, and error-propagation models, and demonstrates the method on heat, wave, Klein-Gordon, and fractional heat equations including a two-dimensional disk geometry.

Core claim

The central mechanism is the endpoint contraction identity for Jacobi polynomials, specialized to Legendre and Chebyshev-T bases where all polynomials evaluate to one at the right endpoint. When a space-time block's solution is stored as coefficients u_{j,k} in a tensor product of temporal and spatial polynomial bases, the final-time spatial profile is obtained by summing over the time index j for each spatial index k. This summation is exact, costs O(N_t times N_x) operations, and requires only O(N_x) storage for the passed vector. The identity also extends to endpoint derivatives via related Jacobi weight formulae, enabling the handling of higher-order-in-time equations, though the paper's

What carries the argument

Jacobi endpoint value identity (Lemma 3.1): shifted Jacobi polynomials evaluated at the right endpoint of a block equal (a+1)_j / j!, independent of block index and step size. For Legendre and Chebyshev-T, this weight is identically one, reducing peel-and-pass to plain coefficient summation. Jacobi endpoint derivative weights (Lemma 3.3) provide analogous contractions for time derivatives at block endpoints.

If this is right

  • Long-time spectral simulations that were previously infeasible due to memory scaling of a single global space-time tensor become tractable, since resident memory is reduced to that of one block.
  • The peel-and-pass mechanism is agnostic to the spatial discretization, so it composes with any sparse orthogonal-polynomial basis on nontrivial geometries (disks, annuli, triangles, etc.) without modification.
  • For problems with complex-time singularities, blocking improves the spectral convergence rate by moving singularities farther away in each block's reference coordinate, an effect the paper confirms empirically.
  • Reusable factorization of the block operator across uniform blocks yields solve-cost ratios that scale favorably compared to a single global solve, especially for dense or high-fill factorizations.

Load-bearing premise

The error propagation bound assumes a uniform per-block stability factor K bounding how much incoming error amplifies to outgoing error. For dissipative problems K is at most one, but for non-dissipative problems like the wave equation, K can be slightly above one, meaning errors grow exponentially as K to the power of the number of blocks over very long horizons.

What would settle it

If the per-block stability factor K grows with the number of blocks or with block count for non-dissipative problems, the accumulated error K^L would grow without bound, making the method unreliable for long-time wave-type simulations.

read the original abstract

Global space-time spectral methods give spectral accuracy in time but typically require the whole space-time history to be resolved and stored on a single tensor-product domain $T \times \Omega$. We record that in an endpoint-benign Legendre or Chebyshev-$T$ time basis, whose polynomials all equal one at the right endpoint, the final time slice of a space-time block is recovered exactly by summing the stored coefficients along the time index. This peel-and-pass step is a special case of a Jacobi endpoint identity, which also gives derivative formulae for higher-order equations. Writing such higher-order equations as first-order systems preserves the benign value-passing structure. The result is a sparse space-time spectral element method that advances block by block, stores only one block, and needs far fewer time coefficients per solve for long-time problems. We prove the identities, give resident-memory, solve-cost and error-propagation models, and demonstrate the method on $(1{+}1)$D heat, wave and Klein--Gordon equations, and on $(2{+}1)$D fractional heat on the disk with weighted Zernike polynomials in space.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 5 minor

Summary. The paper records that in endpoint-benign Legendre or Chebyshev-T time bases, the final time slice of a space-time spectral block is recovered exactly by summing stored coefficients along the time index. This 'peel and pass' step enables block-by-block time stepping for sparse space-time spectral methods, reducing resident memory to a single block while preserving spectral-in-time accuracy. The authors prove the underlying Jacobi endpoint identities (Lemmas 3.1, 3.3), derive memory, cost, and error propagation models (Section 4), and demonstrate the method on heat, wave, Klein-Gordon, and fractional disk heat equations.

Significance. The central identity (Lemma 3.1, Corollary 3.2) is a correct and clean application of classical Jacobi polynomial endpoint values (DLMF 18.6.1). While the individual mathematical ingredients are standard, their specific use as an exact, quadrature-free, re-expansion-free interface mechanism for sparse space-time spectral blocks is a useful contribution that composes with arbitrary sparse spatial discretizations. The paper provides reproducible code, falsifiable error propagation estimates (Theorem 4.5), and honest reporting of empirical stability factors including K>1 for non-dissipative problems. The deep-time Klein-Gordon experiment (Figure 8) directly demonstrates bounded error accumulation over 1227 periods, supporting the practical utility claim.

minor comments (5)
  1. §5.5, Figure 9: The disk fractional heat timing excludes matrix assembly, which is acknowledged but could be more prominently flagged in Table 1 or the figure caption, as it affects the timing comparison.
  2. Theorem 4.5, assumption (3): The spatial error term epsilon_x is assumed bounded per block, and the accumulated bound gives L*epsilon_x. The paper notes (end of §4.3) that spatial error must be controlled separately, but a brief remark on how this is ensured in the experiments (e.g., fixed N_x chosen to under-resolve spatially relative to temporal) would strengthen the connection between theory and practice.
  3. §5.4: The statement that a global solve at N_t=256 remained at O(1) error for the deep-time Klein-Gordon problem is interesting but could benefit from a brief explanation of why (e.g., insufficient resolution for 1227 periods).
  4. Table 1: The deep-time Klein-Gordon row has empty global entries. A footnote or remark explaining that the global solve is infeasible at the required resolution would improve clarity for readers scanning the table.
  5. Lemma 3.3: The derivative weights for Chebyshev-T (Corollary 3.4) involve a product formula; a reference or brief derivation of the simplification from the Jacobi ratio to the double-factorial form would aid readers wishing to verify.

Simulated Author's Rebuttal

5 responses · 0 unresolved

We thank the referee for a careful and positive assessment. The referee's recommendation is minor revision with no major comments listed. We address the referee's summary and significance assessment below, noting where we agree refinements are warranted and where we believe the manuscript already substantiates the claims made.

read point-by-point responses

  1. Referee: The referee's summary describes the paper as recording that endpoint-benign Legendre or Chebyshev-T time bases allow exact recovery of the final time slice by summing coefficients along the time index, enabling block-by-block time stepping with reduced memory and spectral-in-time accuracy.

    Authors: We confirm that this summary accurately characterizes the paper's central contribution. No revision is needed to align the manuscript with this description; the abstract and Section 3 already state this precisely. revision: no

  2. Referee: The referee's significance assessment notes that the central identity (Lemma 3.1, Corollary 3.2) is a correct and clean application of classical Jacobi polynomial endpoint values (DLMF 18.6.1), and that while individual ingredients are standard, their specific use as an exact, quadrature-free, re-expansion-free interface mechanism is a useful contribution composing with arbitrary sparse spatial discretizations.

    Authors: We agree with this characterization. The manuscript already states in Section 3 that the endpoint values are classical and that the contribution lies in their use as the interface mechanism. We will add a brief clarifying sentence in the introduction to make this framing more visible to readers who skim past Section 3. revision: partial

  3. Referee: The referee notes the paper provides reproducible code, falsifiable error propagation estimates (Theorem 4.5), and honest reporting of empirical stability factors including K>1 for non-dissipative problems.

    Authors: We appreciate this assessment. The code repository is archived at the DOI cited in the manuscript. Theorem 4.5 and the empirical K values in Figure 8 (K≈1.058 for L=50 and K≈1.021 for L=100) are reported as stated. No revision needed. revision: no

  4. Referee: The referee highlights the deep-time Klein-Gordon experiment (Figure 8) as directly demonstrating bounded error accumulation over 1227 periods, supporting the practical utility claim.

    Authors: We agree. Figure 8 shows passed-slice error below 4×10⁻¹¹ throughout and below 10⁻¹¹ for L=100, consistent with Theorem 4.5's prediction of mild accumulation when K is close to 1. No revision needed. revision: no

  5. Referee: The referee's recommendation is minor revision, with no major comments enumerated in the report.

    Authors: Since no specific major comments requiring changes were listed, we interpret the minor revision recommendation as an endorsement subject to routine polishing. We will conduct a final proofreading pass for typographical and formatting consistency before resubmission. revision: partial

Circularity Check

0 steps flagged

No significant circularity detected.

full rationale

The paper's central mathematical claim (Lemma 3.1, Corollary 3.2) is a direct, self-contained application of the classical Jacobi polynomial endpoint value P_j^{(a,b)}(1) = binom(j+a, j) (DLMF Table 18.6.1), an external reference. The peel-and-pass identity is derived from first principles by evaluating the polynomial expansion at the right endpoint, with no parameter fitting to target outputs. The error propagation model (Theorem 4.5) is a standard geometric recursion using a per-block stability factor K that is defined independently of the peel-and-pass mechanism itself. Self-citations (references [12], [36], [37]) are used only for external tools (Zernike bases, fractional Laplacian formulae) and do not define the target result. The numerical experiments validate the method against manufactured exact solutions rather than fitting and re-predicting the same data. No step in the derivation chain reduces to its own inputs by construction.

Axiom & Free-Parameter Ledger

3 free parameters · 4 axioms · 0 invented entities

The paper introduces no new physical entities or postulated objects. The free parameters are standard discretization choices (block count, polynomial degrees). The axioms are standard mathematical facts or standard domain assumptions for spectral methods.

free parameters (3)
  • Block count L = varies (e.g., 2, 4, 50, 100)
    Number of uniform time blocks, chosen by the user to trade off block size against accumulation.
  • Time coefficients per block N_t = varies (e.g., 16, 24, 96)
    Polynomial truncation degree in time per block, chosen to reach desired accuracy.
  • Spatial coefficients N_x = varies (e.g., 20, 24, 32)
    Polynomial truncation degree in space, chosen to resolve spatial profile.
axioms (4)
  • domain assumption Existence of sparse orthogonal polynomial bases and operators on the spatial domain.
    Section 2.1 assumes availability of a complete orthogonal-polynomial basis with sparse differentiation and conversion operators, citing the ultraspherical spectral method.
  • domain assumption Exact solution is analytic in a complex neighbourhood of the time interval.
    Theorem 4.5 assumption 1 requires analyticity for the geometric convergence rate of the polynomial approximation.
  • domain assumption Per-block stability factor K bounds outgoing error from incoming error.
    Theorem 4.5 assumption 4 requires a uniform stability factor K for the error propagation bound.
  • standard math DLMF formulas for Jacobi polynomial endpoint values and derivatives.
    Lemmas 3.1 and 3.3 rely on DLMF Table 18.6.1 and 18.9.15 for the underlying polynomial identities.

pith-pipeline@v1.1.0-glm · 18182 in / 2376 out tokens · 240113 ms · 2026-07-08T05:12:50.645629+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

36 extracted references · 36 canonical work pages · 2 internal anchors

  1. [1]

    SIAM Rev.55(3), 462–489 (2013) https://doi.org/10.1137/120865458

    Olver, S., Townsend, A.: A Fast and Well-Conditioned Spectral Method. SIAM Rev.55(3), 462–489 (2013) https://doi.org/10.1137/120865458

    work page doi:10.1137/120865458 2013

  2. [2]

    Acta Numerica29, 573–699 (2020) https://doi.org/10.1017/ S0962492920000045

    Olver, S., Slevinsky, R.M., Townsend, A.: Fast algorithms using orthogo- nal polynomials. Acta Numerica29, 573–699 (2020) https://doi.org/10.1017/ S0962492920000045

    work page 2020

  3. [3]

    Olver, S., Townsend, A., Vasil, G.: A Sparse Spectral Method on Triangles. SIAM J. Sci. Comput.41(6), 3728–3756 (2019) https://doi.org/10.1137/19M1245888

    work page doi:10.1137/19m1245888 2019

  4. [4]

    Vasil, G.M., Burns, K.J., Lecoanet, D., Olver, S., Brown, B.P., Oishi, J.S.: Tensor Calculus in Polar Coordinates Using Jacobi Polynomials. J. Comput. Phys.325, 53–73 (2016) https://doi.org/10.1016/j.jcp.2016.08.013

    work page doi:10.1016/j.jcp.2016.08.013 2016

  5. [5]

    Snowball, B., Olver, S.: Sparse Spectral and p-Finite Element Methods for Partial Differential Equations on Disk Slices and Trapeziums. Stud. Appl. Math.145(1), 3–35 (2020) https://doi.org/10.1111/sapm.12303

    work page doi:10.1111/sapm.12303 2020

  6. [6]

    Snowball, B., Olver, S.: Sparse Spectral Methods for Partial Differential Equations on Spherical Caps. Trans. Math. Appl.5(1), 001 (2021) https://doi.org/10.1093/ imatrm/tnab001

    work page 2021

  7. [7]

    Fasondini, M., Olver, S., Xu, Y.: Orthogonal Polynomials on a Class of Planar Algebraic Curves. Stud. Appl. Math.151(1), 369–405 (2023) https://doi.org/10. 1111/sapm.12582

    work page 2023

  8. [8]

    Foundations of Computational Mathematics23(1), 1–31 (2023) https://doi.org/ 10.1007/s10208-021-09540-w 21

    Fasondini, M., Olver, S., Xu, Y.: Orthogonal polynomials on planar cubic curves. Foundations of Computational Mathematics23(1), 1–31 (2023) https://doi.org/ 10.1007/s10208-021-09540-w 21

    work page doi:10.1007/s10208-021-09540-w 2023

  9. [9]

    A sparse spectral method on a class of domains bounded by planar algebraic curves

    Yao, J., Fasondini, M., Olver, S.: A sparse spectral method on a class of domains bounded by planar algebraic curves. arXiv:2606.24423 (2026)

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  10. [10]

    Papadopoulos, I., Gutleb, T.S., Slevinsky, R.M., Olver, S.: Building Hierarchies of Semiclassical Jacobi Polynomials for Spectral Methods in Annuli. SIAM J. Sci. Comput.46(6), 3448–3476 (2024) https://doi.org/10.1137/23M160846X

    work page doi:10.1137/23m160846x 2024

  11. [11]

    Papadopoulos, I.P.A., Olver, S.: A Sparse Hierarchical hp-Finite Element Method on Disks and Annuli. J. Sci. Comput.104(2), 51 (2025) https://doi.org/10.1007/ s10915-025-02964-4

    work page 2025

  12. [12]

    Gutleb, T.S., Olver, S., Slevinsky, R.M.: Polynomial and Rational Measure Mod- ifications of Orthogonal Polynomials via Infinite-Dimensional Banded Matrix Factorizations. Found. Comput. Math.25(5), 1463–1505 (2025) https://doi.org/ 10.1007/s10208-024-09671-w

    work page doi:10.1007/s10208-024-09671-w 2025

  13. [13]

    Gumerov, K., Rigg, S., Slevinsky, R.M.: Fast Measure Modification of Orthogonal Polynomials via Matrices with Displacement or Hierarchical Off-Diagonal Low- Rank Structure. SIAM J. Sci. Comput.48(2), 624–645 (2026) https://doi.org/ 10.1137/25M1724213

    work page doi:10.1137/25m1724213 2026

  14. [14]

    Journal of Computational Physics436, 110087 (2021) https://doi.org/ 10.1016/j.jcp.2020.110087

    Fortunato, D., Hale, N., Townsend, A.: The ultraspherical spectral element method. Journal of Computational Physics436, 110087 (2021) https://doi.org/ 10.1016/j.jcp.2020.110087

    work page doi:10.1016/j.jcp.2020.110087 2021

  15. [15]

    Cambridge Monographs on Applied and Computational Mathemat- ics

    Hesthaven, J.S., Gottlieb, S., Gottlieb, D.: Spectral Methods for Time-Dependent Problems. Cambridge Monographs on Applied and Computational Mathemat- ics. Cambridge University Press, Cambridge (2007). https://doi.org/10.1017/ CBO9780511618352

    work page 2007

  16. [16]

    Cheng, L., Xu, K.: Solving Time-Dependent PDEs with the Ultraspherical Spectral Method. J. Sci. Comput.96(3), 70 (2023) https://doi.org/10.1007/ s10915-023-02287-2

    work page 2023

  17. [17]

    Tal-Ezer, H.: Spectral Methods in Time for Hyperbolic Equations. SIAM J. Numer. Anal.23(1), 11–26 (1986) https://doi.org/10.1137/0723002

    work page doi:10.1137/0723002 1986

  18. [18]

    Tal-Ezer, H.: Spectral Methods in Time for Parabolic Problems. SIAM J. Numer. Anal.26(1), 1–11 (1989) https://doi.org/10.1137/0726001

    work page doi:10.1137/0726001 1989

  19. [19]

    Ierley, G., Spencer, B., Worthing, R.: Spectral Methods in Time for a Class of Parabolic Partial Differential Equations. J. Comput. Phys.102(1), 88–97 (1992) https://doi.org/10.1016/S0021-9991(05)80008-7

    work page doi:10.1016/s0021-9991(05)80008-7 1992

  20. [20]

    Tang, J.G., Ma, H.P.: Single and Multi-Interval Legendreτ-Methods in Time for Parabolic Equations. Adv. Comput. Math.17(4), 349–367 (2002) https://doi. org/10.1023/A:1016273820035 22

    work page doi:10.1023/a:1016273820035 2002

  21. [21]

    Scheffel, J., Lindvall, K.: Optimizing Time-Spectral Solution of Initial-Value Problems. Am. J. Comput. Math.8(1), 7–26 (2018) https://doi.org/10.4236/ ajcm.2018.81002

    work page arXiv 2018

  22. [22]

    Lui, S.H.: Legendre Spectral Collocation in Space and Time for PDEs. Numer. Math.136(1), 75–99 (2017) https://doi.org/10.1007/s00211-016-0834-x

    work page doi:10.1007/s00211-016-0834-x 2017

  23. [23]

    Electron

    Lui, S.H., Nataj, S.: Chebyshev Spectral Collocation in Space and Time for the Heat Equation. Electron. Trans. Numer. Anal.52, 295–319 (2020) https://doi. org/10.1553/etna vol52s295

    work page doi:10.1553/etna 2020

  24. [24]

    Lui, S.H., Nataj, S.: Spectral Collocation in Space and Time for Linear PDEs. J. Comput. Phys.424, 109843 (2021) https://doi.org/10.1016/j.jcp.2020.109843

    work page doi:10.1016/j.jcp.2020.109843 2021

  25. [25]

    Doc- toral thesis, University of Manitoba (August 2022)

    Kaur, A.: Space-Time Spectral Methods for Partial Differential Equations. Doc- toral thesis, University of Manitoba (August 2022). http://hdl.handle.net/1993/ 36715

    work page 2022

  26. [26]

    Kaur, A., Lui, S.H.: Space-Time Spectral Method for the Stokes Problem. Appl. Numer. Math.187, 206–234 (2023) https://doi.org/10.1016/j.apnum.2023.02. 013

    work page doi:10.1016/j.apnum.2023.02 2023

  27. [27]

    In: Spectral and High-Order Methods for Partial Differential Equations ICOSAHOM 2023

    Kaur, A., Lui, S., Nataj, S., Wilegoda Liyanage, C.P.: Space-Time Spectral Methods for Linear and Nonlinear PDEs. In: Spectral and High-Order Methods for Partial Differential Equations ICOSAHOM 2023. Lecture Notes in Compu- tational Science and Engineering, pp. 89–112. Springer, Cham (2025). https: //doi.org/10.1007/978-3-031-76988-7 4

    work page doi:10.1007/978-3-031-76988-7 2023

  28. [28]

    Pu, T., Fasondini, M.: The Numerical Solution of Fractional Integral Equations via Orthogonal Polynomials in Fractional Powers. Adv. Comput. Math.49(1), 7 (2023) https://doi.org/10.1007/s10444-022-10009-9 [29]NIST Digital Library of Mathematical Functions. https:// dlmf.nist.gov/ , Release 1.2.7 of 2026-06-15. F. W. J. Olver, A. B. Olde Daalhuis, D. W. Lo...

    work page doi:10.1007/s10444-022-10009-9 2023

  29. [29]

    SIAM, Philadelphia (2013)

    Trefethen, L.N.: Approximation Theory and Approximation Practice. SIAM, Philadelphia (2013)

    work page 2013

  30. [30]

    SIAM Rev.59(1), 65–98 (2017) https://doi.org/10.1137/ 141000671

    Bezanson, J., Edelman, A., Karpinski, S., Shah, V.B.: Julia: A Fresh Approach to Numerical Computing. SIAM Rev.59(1), 65–98 (2017) https://doi.org/10.1137/ 141000671

    work page 2017

  31. [31]

    Accessed 17 June 2026 (2026)

    JuliaApproximation: ClassicalOrthogonalPolynomials.jl. Accessed 17 June 2026 (2026). https://github.com/JuliaApproximation/ ClassicalOrthogonalPolynomials.jl 23

    work page 2026

  32. [32]

    Accessed 17 June 2026 (2026)

    JuliaApproximation: MultivariateOrthogonalPolynomials.jl. Accessed 17 June 2026 (2026). https://github.com/JuliaApproximation/ MultivariateOrthogonalPolynomials.jl

    work page 2026

  33. [33]

    Zenodo (2026)

    Gutleb, T.S.: TSGut/SpacetimeSpectralPassExamples.jl: v0.1.0. Zenodo (2026). https://doi.org/10.5281/zenodo.20938999

    work page doi:10.5281/zenodo.20938999 2026

  34. [34]

    Kwa´ snicki, M.: Ten Equivalent Definitions of the Fractional Laplace Oper- ator. Fract. Calc. Appl. Anal.20(1), 7–51 (2017) https://doi.org/10.1515/ fca-2017-0002

    work page 2017

  35. [35]

    Gutleb, T.S., Carrillo, J.A., Olver, S.: Computation of Power Law Equilibrium Measures on Balls of Arbitrary Dimension. Constr. Approx.58, 75–120 (2023) https://doi.org/10.1007/s00365-022-09606-0

    work page doi:10.1007/s00365-022-09606-0 2023

  36. [36]

    Explicit fractional Laplacians and Riesz potentials of classical functions

    Gutleb, T.S., Papadopoulos, I.P.A.: Explicit Fractional Laplacians and Riesz Potentials of Classical Functions. arXiv:2311.10896 (2023) https://doi.org/10. 48550/arXiv.2311.10896 24

    work page internal anchor Pith review Pith/arXiv arXiv 2023